【摘要】 在地面激光扫描仪的建筑测量中，由于镜面反射和透明玻璃物体产生的反射干扰，测量数据中含有大量噪声点。这些反射噪声会严重影响后续数据分析的准确性。本研究提出了一种高效的反射噪声滤除方法：首先基于点云反射强度估计玻璃区域(即反射噪声源);结合激光扫描仪位置与玻璃区域进行光束路径追踪，以确定可能产生反射噪声的空间范围；最后通过融合基于噪声点云-实体对称性的几何约束和三维特征相似性估计网络的深度特征联合判别机制检测反射噪声，并实现针对性去除。实验结果表明，该方法反射噪声去除率可达90%,能有效实现测量点云的选择性降噪。更多还原

【Abstract】 In terrestrial laser scanning(TLS) measurement for architectural measurement, specular reflections and interference from transparent glass objects introduce significant noise points into the measured data. These reflective noises severely compromise the accuracy of subsequent data analysis. This study proposes an efficient reflective noise filtering method: First, glass regions(i.e.,noise sources) are estimated based on point cloud intensity. Next, beam path tracing is performed by combining the laser scanner’s position and the identified glass regions to delineate spatial ranges prone to reflective noise. Finally, a joint discrimination mechanism integrates geometric constraints derived from reflection symmetry and deep features from a 3D feature similarity estimation network to detect reflective noise, enabling targeted removal. Experimental results demonstrate a 90% reflective noise removal rate, achieving selective denoising of measured point clouds.更多还原